how to avoid for loop using pandas dataframe

Asked Nov 09 '21 at 15:58

Active Nov 09 '21 at 16:06

Viewed 100 times

def moving_window(self, input_df: pd.DataFrame, previous_df: pd.DataFrame):
    comb_df = pd.concat([previous_df, input_df])
    results = pd.Series()
    for index, row in input_df.iterrows():
        start_ts = int(index) - self.duration * 1000
        current_df = comb_df.loc[(comb_df.index >= start_ts) & (comb_df.index <= int(index))]
        single_pred = self.another_function(current_df)
        if single_pred:
            results.loc[index] = single_pred
    return reults

Is there a way to convert the for loop into something more efficient ?

edited Nov 09 '21 at 16:06

asked Nov 09 '21 at 15:58

gammawind

Are you familiar with [`pandas.DataFrame.rolling`](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.rolling.html)? Just judging by your function name `moving_window`, it seems like it would be relevant. – Joshua Voskamp Nov 09 '21 at 16:07
What is the `type` of `self.duration` and `input_df.index`? – Joshua Voskamp Nov 09 '21 at 16:09
i tried to use rolling, but not sure how to set/override window param, basically i want to have set of rows between two timestamps – gammawind Nov 09 '21 at 16:10
@JoshuaVoskamp epoch timestamp(int64) – gammawind Nov 09 '21 at 16:10
1

Have a look at [time offsets](https://pandas.pydata.org/pandas-docs/stable/user_guide/timeseries.html#dateoffset-objects) for using in your `.rolling` – Joshua Voskamp Nov 09 '21 at 16:11
sure ! thanks @JoshuaVoskamp – gammawind Nov 09 '21 at 16:14
1

[this SO answer](https://stackoverflow.com/questions/15771472/pandas-rolling-mean-by-time-interval) to a possibly-related question may help point you in the right direction – Joshua Voskamp Nov 09 '21 at 16:15

how to avoid for loop using pandas dataframe

0 Answers0